Intelligent Control Of Robot Manipulators With Unknown Model

The control problems of robot have received great attention in theoretical and engineering for many years. When the robot model is exactly known, the technique of feedback linearization in nonlinear systems can solve the problems very well. However, the parameters of dynamic model of robotic manipulators may also be subject to change when the manipulator goes about its task. Meanwhile, the system can be influenced by uncertainties such as external disturbance and payload change. Therefore, it is necessary to propose other control methods so as to enhance the adaptive ability and robustness.In this dissertation, the robotic system with unknown model is regarded as controlled plant. The various position and force control schemes based intelligent algorithms are developed using the references available.The dissertation gives a brief description about the developing situation and control method firstly, and then the models of robot and mathematics knowledge are introduced in detail. After that, a RBF neural network for robotic trajectory tracking is firstly presented, which adopts neural network to establish the model of robot. The control scheme combines H_∞ control theory and RBF neural network algorithm organically to eliminate approximation errors and restrain disturbances. Then for the benefit of Cartesian space' control, an adaptive fuzzy control of robotic manipulators with uncertainties in task space is proposed, which does not require the inverse of the Jacobian matrix. And the proposed controller does not contain the inverse of matrices and can do without singularities.It is necessary to control not only the position of a manipulator but also the force exerted by the end-effector on an object, when robot manipulators contact with the environment. Therefore, a fuzzy-neuro position/force control of robot manipulators is finally presented, which can achieve the position/force control without changing the inner position controller. The proposed methoduses a suitable neural network to directly parameterize the control law, and a self-tuning fuzzy system though which the contact force can be regulated by appropriate control of the end-effector position. It is proven that the proposed controller has high value in practice.